A hybrid deep neural network based prediction of 300 MW coal-fired boiler combustion operation condition

In power generation industries, boilers are required to be operated under a range of different conditions toaccommodate demands for fuel randomness and energy fluctuation. Reliable prediction of the combustionoperation condition is crucial for an in-depth understanding of boiler performance and maintaining high combustionefficiency. However, it is difficult to establish an accurate prediction model based on traditional data-driven methods,which requires prior expert knowledge and a large number of labeled data. To overcome these limitations, a novelprediction method for the combustion operation condition based on flame imaging and hybrid deep neural networkis proposed. The proposed hybrid model is a combination of convolutional sparse autoencoder (CSAE) and leastsupport vector machine (LSSVM), i.e., CSAE-LSSVM, where the convolutional sparse autoencoder with deeparchitectures is utilized to extract the essential features of flame image, and then essential features are input intothe least support vector machine for operation condition prediction. A comprehensive investigation of optimalhyper-parameter and dropout technique is carried out to improve the performance of the CSAE-LSSVM. Theeffectiveness of the proposed model is evaluated by 300MW tangential coal-fired boiler flame images. Theprediction accuracy of the proposed hybrid model reaches 98.06%, and its prediction time is 3.06millisecond/image. It is observed that the proposed model could present a superior performance in comparison toother existing neural network models.